xref: /freebsd/sys/compat/linux/linux_misc.c (revision eda14cbc264d6969b02f2b1994cef11148e914f1)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2002 Doug Rabson
5  * Copyright (c) 1994-1995 Søren Schmidt
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer
13  *    in this position and unchanged.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. The name of the author may not be used to endorse or promote products
18  *    derived from this software without specific prior written permission
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
40 #if defined(__i386__)
41 #include <sys/imgact_aout.h>
42 #endif
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mman.h>
49 #include <sys/mount.h>
50 #include <sys/msgbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/procctl.h>
56 #include <sys/reboot.h>
57 #include <sys/racct.h>
58 #include <sys/random.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sched.h>
61 #include <sys/sdt.h>
62 #include <sys/signalvar.h>
63 #include <sys/stat.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/systm.h>
68 #include <sys/time.h>
69 #include <sys/vmmeter.h>
70 #include <sys/vnode.h>
71 #include <sys/wait.h>
72 #include <sys/cpuset.h>
73 #include <sys/uio.h>
74 
75 #include <security/mac/mac_framework.h>
76 
77 #include <vm/vm.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_kern.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.h>
82 #include <vm/swap_pager.h>
83 
84 #ifdef COMPAT_LINUX32
85 #include <machine/../linux32/linux.h>
86 #include <machine/../linux32/linux32_proto.h>
87 #else
88 #include <machine/../linux/linux.h>
89 #include <machine/../linux/linux_proto.h>
90 #endif
91 
92 #include <compat/linux/linux_dtrace.h>
93 #include <compat/linux/linux_file.h>
94 #include <compat/linux/linux_mib.h>
95 #include <compat/linux/linux_signal.h>
96 #include <compat/linux/linux_timer.h>
97 #include <compat/linux/linux_util.h>
98 #include <compat/linux/linux_sysproto.h>
99 #include <compat/linux/linux_emul.h>
100 #include <compat/linux/linux_misc.h>
101 
102 /**
103  * Special DTrace provider for the linuxulator.
104  *
105  * In this file we define the provider for the entire linuxulator. All
106  * modules (= files of the linuxulator) use it.
107  *
108  * We define a different name depending on the emulated bitsize, see
109  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
110  *      native bitsize          = linuxulator
111  *      amd64, 32bit emulation  = linuxulator32
112  */
113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
114 
115 int stclohz;				/* Statistics clock frequency */
116 
117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
118 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
119 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
120 	RLIMIT_MEMLOCK, RLIMIT_AS
121 };
122 
123 struct l_sysinfo {
124 	l_long		uptime;		/* Seconds since boot */
125 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
126 #define LINUX_SYSINFO_LOADS_SCALE 65536
127 	l_ulong		totalram;	/* Total usable main memory size */
128 	l_ulong		freeram;	/* Available memory size */
129 	l_ulong		sharedram;	/* Amount of shared memory */
130 	l_ulong		bufferram;	/* Memory used by buffers */
131 	l_ulong		totalswap;	/* Total swap space size */
132 	l_ulong		freeswap;	/* swap space still available */
133 	l_ushort	procs;		/* Number of current processes */
134 	l_ushort	pads;
135 	l_ulong		totalhigh;
136 	l_ulong		freehigh;
137 	l_uint		mem_unit;
138 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
139 };
140 
141 struct l_pselect6arg {
142 	l_uintptr_t	ss;
143 	l_size_t	ss_len;
144 };
145 
146 static int	linux_utimensat_nsec_valid(l_long);
147 
148 
149 int
150 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
151 {
152 	struct l_sysinfo sysinfo;
153 	int i, j;
154 	struct timespec ts;
155 
156 	bzero(&sysinfo, sizeof(sysinfo));
157 	getnanouptime(&ts);
158 	if (ts.tv_nsec != 0)
159 		ts.tv_sec++;
160 	sysinfo.uptime = ts.tv_sec;
161 
162 	/* Use the information from the mib to get our load averages */
163 	for (i = 0; i < 3; i++)
164 		sysinfo.loads[i] = averunnable.ldavg[i] *
165 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
166 
167 	sysinfo.totalram = physmem * PAGE_SIZE;
168 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
169 
170 	/*
171 	 * sharedram counts pages allocated to named, swap-backed objects such
172 	 * as shared memory segments and tmpfs files.  There is no cheap way to
173 	 * compute this, so just leave the field unpopulated.  Linux itself only
174 	 * started setting this field in the 3.x timeframe.
175 	 */
176 	sysinfo.sharedram = 0;
177 	sysinfo.bufferram = 0;
178 
179 	swap_pager_status(&i, &j);
180 	sysinfo.totalswap = i * PAGE_SIZE;
181 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
182 
183 	sysinfo.procs = nprocs;
184 
185 	/*
186 	 * Platforms supported by the emulation layer do not have a notion of
187 	 * high memory.
188 	 */
189 	sysinfo.totalhigh = 0;
190 	sysinfo.freehigh = 0;
191 
192 	sysinfo.mem_unit = 1;
193 
194 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
195 }
196 
197 #ifdef LINUX_LEGACY_SYSCALLS
198 int
199 linux_alarm(struct thread *td, struct linux_alarm_args *args)
200 {
201 	struct itimerval it, old_it;
202 	u_int secs;
203 	int error;
204 
205 	secs = args->secs;
206 	/*
207 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
208 	 * to match kern_setitimer()'s limit to avoid error from it.
209 	 *
210 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
211 	 * platforms.
212 	 */
213 	if (secs > INT32_MAX / 2)
214 		secs = INT32_MAX / 2;
215 
216 	it.it_value.tv_sec = secs;
217 	it.it_value.tv_usec = 0;
218 	timevalclear(&it.it_interval);
219 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
220 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
221 
222 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
223 	    old_it.it_value.tv_usec >= 500000)
224 		old_it.it_value.tv_sec++;
225 	td->td_retval[0] = old_it.it_value.tv_sec;
226 	return (0);
227 }
228 #endif
229 
230 int
231 linux_brk(struct thread *td, struct linux_brk_args *args)
232 {
233 	struct vmspace *vm = td->td_proc->p_vmspace;
234 	uintptr_t new, old;
235 
236 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
237 	new = (uintptr_t)args->dsend;
238 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
239 		td->td_retval[0] = (register_t)new;
240 	else
241 		td->td_retval[0] = (register_t)old;
242 
243 	return (0);
244 }
245 
246 #if defined(__i386__)
247 /* XXX: what about amd64/linux32? */
248 
249 int
250 linux_uselib(struct thread *td, struct linux_uselib_args *args)
251 {
252 	struct nameidata ni;
253 	struct vnode *vp;
254 	struct exec *a_out;
255 	vm_map_t map;
256 	vm_map_entry_t entry;
257 	struct vattr attr;
258 	vm_offset_t vmaddr;
259 	unsigned long file_offset;
260 	unsigned long bss_size;
261 	char *library;
262 	ssize_t aresid;
263 	int error;
264 	bool locked, opened, textset;
265 
266 	a_out = NULL;
267 	vp = NULL;
268 	locked = false;
269 	textset = false;
270 	opened = false;
271 
272 	if (!LUSECONVPATH(td)) {
273 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
274 		    UIO_USERSPACE, args->library, td);
275 		error = namei(&ni);
276 	} else {
277 		LCONVPATHEXIST(td, args->library, &library);
278 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
279 		    UIO_SYSSPACE, library, td);
280 		error = namei(&ni);
281 		LFREEPATH(library);
282 	}
283 	if (error)
284 		goto cleanup;
285 
286 	vp = ni.ni_vp;
287 	NDFREE(&ni, NDF_ONLY_PNBUF);
288 
289 	/*
290 	 * From here on down, we have a locked vnode that must be unlocked.
291 	 * XXX: The code below largely duplicates exec_check_permissions().
292 	 */
293 	locked = true;
294 
295 	/* Executable? */
296 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
297 	if (error)
298 		goto cleanup;
299 
300 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
301 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
302 		/* EACCESS is what exec(2) returns. */
303 		error = ENOEXEC;
304 		goto cleanup;
305 	}
306 
307 	/* Sensible size? */
308 	if (attr.va_size == 0) {
309 		error = ENOEXEC;
310 		goto cleanup;
311 	}
312 
313 	/* Can we access it? */
314 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
315 	if (error)
316 		goto cleanup;
317 
318 	/*
319 	 * XXX: This should use vn_open() so that it is properly authorized,
320 	 * and to reduce code redundancy all over the place here.
321 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
322 	 * than vn_open().
323 	 */
324 #ifdef MAC
325 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
326 	if (error)
327 		goto cleanup;
328 #endif
329 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
330 	if (error)
331 		goto cleanup;
332 	opened = true;
333 
334 	/* Pull in executable header into exec_map */
335 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
336 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
337 	if (error)
338 		goto cleanup;
339 
340 	/* Is it a Linux binary ? */
341 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
342 		error = ENOEXEC;
343 		goto cleanup;
344 	}
345 
346 	/*
347 	 * While we are here, we should REALLY do some more checks
348 	 */
349 
350 	/* Set file/virtual offset based on a.out variant. */
351 	switch ((int)(a_out->a_magic & 0xffff)) {
352 	case 0413:			/* ZMAGIC */
353 		file_offset = 1024;
354 		break;
355 	case 0314:			/* QMAGIC */
356 		file_offset = 0;
357 		break;
358 	default:
359 		error = ENOEXEC;
360 		goto cleanup;
361 	}
362 
363 	bss_size = round_page(a_out->a_bss);
364 
365 	/* Check various fields in header for validity/bounds. */
366 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
367 		error = ENOEXEC;
368 		goto cleanup;
369 	}
370 
371 	/* text + data can't exceed file size */
372 	if (a_out->a_data + a_out->a_text > attr.va_size) {
373 		error = EFAULT;
374 		goto cleanup;
375 	}
376 
377 	/*
378 	 * text/data/bss must not exceed limits
379 	 * XXX - this is not complete. it should check current usage PLUS
380 	 * the resources needed by this library.
381 	 */
382 	PROC_LOCK(td->td_proc);
383 	if (a_out->a_text > maxtsiz ||
384 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
385 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
386 	    bss_size) != 0) {
387 		PROC_UNLOCK(td->td_proc);
388 		error = ENOMEM;
389 		goto cleanup;
390 	}
391 	PROC_UNLOCK(td->td_proc);
392 
393 	/*
394 	 * Prevent more writers.
395 	 */
396 	error = VOP_SET_TEXT(vp);
397 	if (error != 0)
398 		goto cleanup;
399 	textset = true;
400 
401 	/*
402 	 * Lock no longer needed
403 	 */
404 	locked = false;
405 	VOP_UNLOCK(vp);
406 
407 	/*
408 	 * Check if file_offset page aligned. Currently we cannot handle
409 	 * misalinged file offsets, and so we read in the entire image
410 	 * (what a waste).
411 	 */
412 	if (file_offset & PAGE_MASK) {
413 		/* Map text+data read/write/execute */
414 
415 		/* a_entry is the load address and is page aligned */
416 		vmaddr = trunc_page(a_out->a_entry);
417 
418 		/* get anon user mapping, read+write+execute */
419 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
420 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
421 		    VM_PROT_ALL, VM_PROT_ALL, 0);
422 		if (error)
423 			goto cleanup;
424 
425 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
426 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
427 		    td->td_ucred, NOCRED, &aresid, td);
428 		if (error != 0)
429 			goto cleanup;
430 		if (aresid != 0) {
431 			error = ENOEXEC;
432 			goto cleanup;
433 		}
434 	} else {
435 		/*
436 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
437 		 * to skip the executable header
438 		 */
439 		vmaddr = trunc_page(a_out->a_entry);
440 
441 		/*
442 		 * Map it all into the process's space as a single
443 		 * copy-on-write "data" segment.
444 		 */
445 		map = &td->td_proc->p_vmspace->vm_map;
446 		error = vm_mmap(map, &vmaddr,
447 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
448 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
449 		if (error)
450 			goto cleanup;
451 		vm_map_lock(map);
452 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
453 			vm_map_unlock(map);
454 			error = EDOOFUS;
455 			goto cleanup;
456 		}
457 		entry->eflags |= MAP_ENTRY_VN_EXEC;
458 		vm_map_unlock(map);
459 		textset = false;
460 	}
461 
462 	if (bss_size != 0) {
463 		/* Calculate BSS start address */
464 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
465 		    a_out->a_data;
466 
467 		/* allocate some 'anon' space */
468 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
469 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
470 		    VM_PROT_ALL, 0);
471 		if (error)
472 			goto cleanup;
473 	}
474 
475 cleanup:
476 	if (opened) {
477 		if (locked)
478 			VOP_UNLOCK(vp);
479 		locked = false;
480 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
481 	}
482 	if (textset) {
483 		if (!locked) {
484 			locked = true;
485 			VOP_LOCK(vp, LK_SHARED | LK_RETRY);
486 		}
487 		VOP_UNSET_TEXT_CHECKED(vp);
488 	}
489 	if (locked)
490 		VOP_UNLOCK(vp);
491 
492 	/* Release the temporary mapping. */
493 	if (a_out)
494 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
495 
496 	return (error);
497 }
498 
499 #endif	/* __i386__ */
500 
501 #ifdef LINUX_LEGACY_SYSCALLS
502 int
503 linux_select(struct thread *td, struct linux_select_args *args)
504 {
505 	l_timeval ltv;
506 	struct timeval tv0, tv1, utv, *tvp;
507 	int error;
508 
509 	/*
510 	 * Store current time for computation of the amount of
511 	 * time left.
512 	 */
513 	if (args->timeout) {
514 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
515 			goto select_out;
516 		utv.tv_sec = ltv.tv_sec;
517 		utv.tv_usec = ltv.tv_usec;
518 
519 		if (itimerfix(&utv)) {
520 			/*
521 			 * The timeval was invalid.  Convert it to something
522 			 * valid that will act as it does under Linux.
523 			 */
524 			utv.tv_sec += utv.tv_usec / 1000000;
525 			utv.tv_usec %= 1000000;
526 			if (utv.tv_usec < 0) {
527 				utv.tv_sec -= 1;
528 				utv.tv_usec += 1000000;
529 			}
530 			if (utv.tv_sec < 0)
531 				timevalclear(&utv);
532 		}
533 		microtime(&tv0);
534 		tvp = &utv;
535 	} else
536 		tvp = NULL;
537 
538 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
539 	    args->exceptfds, tvp, LINUX_NFDBITS);
540 	if (error)
541 		goto select_out;
542 
543 	if (args->timeout) {
544 		if (td->td_retval[0]) {
545 			/*
546 			 * Compute how much time was left of the timeout,
547 			 * by subtracting the current time and the time
548 			 * before we started the call, and subtracting
549 			 * that result from the user-supplied value.
550 			 */
551 			microtime(&tv1);
552 			timevalsub(&tv1, &tv0);
553 			timevalsub(&utv, &tv1);
554 			if (utv.tv_sec < 0)
555 				timevalclear(&utv);
556 		} else
557 			timevalclear(&utv);
558 		ltv.tv_sec = utv.tv_sec;
559 		ltv.tv_usec = utv.tv_usec;
560 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
561 			goto select_out;
562 	}
563 
564 select_out:
565 	return (error);
566 }
567 #endif
568 
569 int
570 linux_mremap(struct thread *td, struct linux_mremap_args *args)
571 {
572 	uintptr_t addr;
573 	size_t len;
574 	int error = 0;
575 
576 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
577 		td->td_retval[0] = 0;
578 		return (EINVAL);
579 	}
580 
581 	/*
582 	 * Check for the page alignment.
583 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
584 	 */
585 	if (args->addr & PAGE_MASK) {
586 		td->td_retval[0] = 0;
587 		return (EINVAL);
588 	}
589 
590 	args->new_len = round_page(args->new_len);
591 	args->old_len = round_page(args->old_len);
592 
593 	if (args->new_len > args->old_len) {
594 		td->td_retval[0] = 0;
595 		return (ENOMEM);
596 	}
597 
598 	if (args->new_len < args->old_len) {
599 		addr = args->addr + args->new_len;
600 		len = args->old_len - args->new_len;
601 		error = kern_munmap(td, addr, len);
602 	}
603 
604 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
605 	return (error);
606 }
607 
608 #define LINUX_MS_ASYNC       0x0001
609 #define LINUX_MS_INVALIDATE  0x0002
610 #define LINUX_MS_SYNC        0x0004
611 
612 int
613 linux_msync(struct thread *td, struct linux_msync_args *args)
614 {
615 
616 	return (kern_msync(td, args->addr, args->len,
617 	    args->fl & ~LINUX_MS_SYNC));
618 }
619 
620 #ifdef LINUX_LEGACY_SYSCALLS
621 int
622 linux_time(struct thread *td, struct linux_time_args *args)
623 {
624 	struct timeval tv;
625 	l_time_t tm;
626 	int error;
627 
628 	microtime(&tv);
629 	tm = tv.tv_sec;
630 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
631 		return (error);
632 	td->td_retval[0] = tm;
633 	return (0);
634 }
635 #endif
636 
637 struct l_times_argv {
638 	l_clock_t	tms_utime;
639 	l_clock_t	tms_stime;
640 	l_clock_t	tms_cutime;
641 	l_clock_t	tms_cstime;
642 };
643 
644 
645 /*
646  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
647  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
648  * auxiliary vector entry.
649  */
650 #define	CLK_TCK		100
651 
652 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
653 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
654 
655 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
656 			    CONVNTCK(r) : CONVOTCK(r))
657 
658 int
659 linux_times(struct thread *td, struct linux_times_args *args)
660 {
661 	struct timeval tv, utime, stime, cutime, cstime;
662 	struct l_times_argv tms;
663 	struct proc *p;
664 	int error;
665 
666 	if (args->buf != NULL) {
667 		p = td->td_proc;
668 		PROC_LOCK(p);
669 		PROC_STATLOCK(p);
670 		calcru(p, &utime, &stime);
671 		PROC_STATUNLOCK(p);
672 		calccru(p, &cutime, &cstime);
673 		PROC_UNLOCK(p);
674 
675 		tms.tms_utime = CONVTCK(utime);
676 		tms.tms_stime = CONVTCK(stime);
677 
678 		tms.tms_cutime = CONVTCK(cutime);
679 		tms.tms_cstime = CONVTCK(cstime);
680 
681 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
682 			return (error);
683 	}
684 
685 	microuptime(&tv);
686 	td->td_retval[0] = (int)CONVTCK(tv);
687 	return (0);
688 }
689 
690 int
691 linux_newuname(struct thread *td, struct linux_newuname_args *args)
692 {
693 	struct l_new_utsname utsname;
694 	char osname[LINUX_MAX_UTSNAME];
695 	char osrelease[LINUX_MAX_UTSNAME];
696 	char *p;
697 
698 	linux_get_osname(td, osname);
699 	linux_get_osrelease(td, osrelease);
700 
701 	bzero(&utsname, sizeof(utsname));
702 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
703 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
704 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
705 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
706 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
707 	for (p = utsname.version; *p != '\0'; ++p)
708 		if (*p == '\n') {
709 			*p = '\0';
710 			break;
711 		}
712 #if defined(__amd64__)
713 	/*
714 	 * On amd64, Linux uname(2) needs to return "x86_64"
715 	 * for both 64-bit and 32-bit applications.  On 32-bit,
716 	 * the string returned by getauxval(AT_PLATFORM) needs
717 	 * to remain "i686", though.
718 	 */
719 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
720 #else
721 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
722 #endif
723 
724 	return (copyout(&utsname, args->buf, sizeof(utsname)));
725 }
726 
727 struct l_utimbuf {
728 	l_time_t l_actime;
729 	l_time_t l_modtime;
730 };
731 
732 #ifdef LINUX_LEGACY_SYSCALLS
733 int
734 linux_utime(struct thread *td, struct linux_utime_args *args)
735 {
736 	struct timeval tv[2], *tvp;
737 	struct l_utimbuf lut;
738 	char *fname;
739 	int error;
740 	bool convpath;
741 
742 	convpath = LUSECONVPATH(td);
743 	if (convpath)
744 		LCONVPATHEXIST(td, args->fname, &fname);
745 
746 	if (args->times) {
747 		if ((error = copyin(args->times, &lut, sizeof lut))) {
748 			if (convpath)
749 				LFREEPATH(fname);
750 			return (error);
751 		}
752 		tv[0].tv_sec = lut.l_actime;
753 		tv[0].tv_usec = 0;
754 		tv[1].tv_sec = lut.l_modtime;
755 		tv[1].tv_usec = 0;
756 		tvp = tv;
757 	} else
758 		tvp = NULL;
759 
760 	if (!convpath) {
761 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
762 		    tvp, UIO_SYSSPACE);
763 	} else {
764 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
765 		    UIO_SYSSPACE);
766 		LFREEPATH(fname);
767 	}
768 	return (error);
769 }
770 #endif
771 
772 #ifdef LINUX_LEGACY_SYSCALLS
773 int
774 linux_utimes(struct thread *td, struct linux_utimes_args *args)
775 {
776 	l_timeval ltv[2];
777 	struct timeval tv[2], *tvp = NULL;
778 	char *fname;
779 	int error;
780 	bool convpath;
781 
782 	convpath = LUSECONVPATH(td);
783 	if (convpath)
784 		LCONVPATHEXIST(td, args->fname, &fname);
785 
786 	if (args->tptr != NULL) {
787 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
788 			LFREEPATH(fname);
789 			return (error);
790 		}
791 		tv[0].tv_sec = ltv[0].tv_sec;
792 		tv[0].tv_usec = ltv[0].tv_usec;
793 		tv[1].tv_sec = ltv[1].tv_sec;
794 		tv[1].tv_usec = ltv[1].tv_usec;
795 		tvp = tv;
796 	}
797 
798 	if (!convpath) {
799 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
800 		    tvp, UIO_SYSSPACE);
801 	} else {
802 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
803 		    tvp, UIO_SYSSPACE);
804 		LFREEPATH(fname);
805 	}
806 	return (error);
807 }
808 #endif
809 
810 static int
811 linux_utimensat_nsec_valid(l_long nsec)
812 {
813 
814 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
815 		return (0);
816 	if (nsec >= 0 && nsec <= 999999999)
817 		return (0);
818 	return (1);
819 }
820 
821 int
822 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
823 {
824 	struct l_timespec l_times[2];
825 	struct timespec times[2], *timesp = NULL;
826 	char *path = NULL;
827 	int error, dfd, flags = 0;
828 
829 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
830 
831 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
832 		return (EINVAL);
833 
834 	if (args->times != NULL) {
835 		error = copyin(args->times, l_times, sizeof(l_times));
836 		if (error != 0)
837 			return (error);
838 
839 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
840 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
841 			return (EINVAL);
842 
843 		times[0].tv_sec = l_times[0].tv_sec;
844 		switch (l_times[0].tv_nsec)
845 		{
846 		case LINUX_UTIME_OMIT:
847 			times[0].tv_nsec = UTIME_OMIT;
848 			break;
849 		case LINUX_UTIME_NOW:
850 			times[0].tv_nsec = UTIME_NOW;
851 			break;
852 		default:
853 			times[0].tv_nsec = l_times[0].tv_nsec;
854 		}
855 
856 		times[1].tv_sec = l_times[1].tv_sec;
857 		switch (l_times[1].tv_nsec)
858 		{
859 		case LINUX_UTIME_OMIT:
860 			times[1].tv_nsec = UTIME_OMIT;
861 			break;
862 		case LINUX_UTIME_NOW:
863 			times[1].tv_nsec = UTIME_NOW;
864 			break;
865 		default:
866 			times[1].tv_nsec = l_times[1].tv_nsec;
867 			break;
868 		}
869 		timesp = times;
870 
871 		/* This breaks POSIX, but is what the Linux kernel does
872 		 * _on purpose_ (documented in the man page for utimensat(2)),
873 		 * so we must follow that behaviour. */
874 		if (times[0].tv_nsec == UTIME_OMIT &&
875 		    times[1].tv_nsec == UTIME_OMIT)
876 			return (0);
877 	}
878 
879 	if (args->pathname != NULL)
880 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
881 	else if (args->flags != 0)
882 		return (EINVAL);
883 
884 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
885 		flags |= AT_SYMLINK_NOFOLLOW;
886 
887 	if (path == NULL)
888 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
889 	else {
890 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
891 			UIO_SYSSPACE, flags);
892 		LFREEPATH(path);
893 	}
894 
895 	return (error);
896 }
897 
898 #ifdef LINUX_LEGACY_SYSCALLS
899 int
900 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
901 {
902 	l_timeval ltv[2];
903 	struct timeval tv[2], *tvp = NULL;
904 	char *fname;
905 	int error, dfd;
906 	bool convpath;
907 
908 	convpath = LUSECONVPATH(td);
909 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
910 	if (convpath)
911 		LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
912 
913 	if (args->utimes != NULL) {
914 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
915 			if (convpath)
916 				LFREEPATH(fname);
917 			return (error);
918 		}
919 		tv[0].tv_sec = ltv[0].tv_sec;
920 		tv[0].tv_usec = ltv[0].tv_usec;
921 		tv[1].tv_sec = ltv[1].tv_sec;
922 		tv[1].tv_usec = ltv[1].tv_usec;
923 		tvp = tv;
924 	}
925 
926 	if (!convpath) {
927 		error = kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
928 		    tvp, UIO_SYSSPACE);
929 	} else {
930 		error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
931 		LFREEPATH(fname);
932 	}
933 	return (error);
934 }
935 #endif
936 
937 static int
938 linux_common_wait(struct thread *td, int pid, int *statusp,
939     int options, struct __wrusage *wrup)
940 {
941 	siginfo_t siginfo;
942 	idtype_t idtype;
943 	id_t id;
944 	int error, status, tmpstat;
945 
946 	if (pid == WAIT_ANY) {
947 		idtype = P_ALL;
948 		id = 0;
949 	} else if (pid < 0) {
950 		idtype = P_PGID;
951 		id = (id_t)-pid;
952 	} else {
953 		idtype = P_PID;
954 		id = (id_t)pid;
955 	}
956 
957 	/*
958 	 * For backward compatibility we implicitly add flags WEXITED
959 	 * and WTRAPPED here.
960 	 */
961 	options |= WEXITED | WTRAPPED;
962 	error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
963 	if (error)
964 		return (error);
965 
966 	if (statusp) {
967 		tmpstat = status & 0xffff;
968 		if (WIFSIGNALED(tmpstat)) {
969 			tmpstat = (tmpstat & 0xffffff80) |
970 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
971 		} else if (WIFSTOPPED(tmpstat)) {
972 			tmpstat = (tmpstat & 0xffff00ff) |
973 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
974 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
975 			if (WSTOPSIG(status) == SIGTRAP) {
976 				tmpstat = linux_ptrace_status(td,
977 				    siginfo.si_pid, tmpstat);
978 			}
979 #endif
980 		} else if (WIFCONTINUED(tmpstat)) {
981 			tmpstat = 0xffff;
982 		}
983 		error = copyout(&tmpstat, statusp, sizeof(int));
984 	}
985 
986 	return (error);
987 }
988 
989 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
990 int
991 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
992 {
993 	struct linux_wait4_args wait4_args;
994 
995 	wait4_args.pid = args->pid;
996 	wait4_args.status = args->status;
997 	wait4_args.options = args->options;
998 	wait4_args.rusage = NULL;
999 
1000 	return (linux_wait4(td, &wait4_args));
1001 }
1002 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1003 
1004 int
1005 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1006 {
1007 	int error, options;
1008 	struct __wrusage wru, *wrup;
1009 
1010 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1011 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1012 		return (EINVAL);
1013 
1014 	options = WEXITED;
1015 	linux_to_bsd_waitopts(args->options, &options);
1016 
1017 	if (args->rusage != NULL)
1018 		wrup = &wru;
1019 	else
1020 		wrup = NULL;
1021 	error = linux_common_wait(td, args->pid, args->status, options, wrup);
1022 	if (error != 0)
1023 		return (error);
1024 	if (args->rusage != NULL)
1025 		error = linux_copyout_rusage(&wru.wru_self, args->rusage);
1026 	return (error);
1027 }
1028 
1029 int
1030 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1031 {
1032 	int status, options, sig;
1033 	struct __wrusage wru;
1034 	siginfo_t siginfo;
1035 	l_siginfo_t lsi;
1036 	idtype_t idtype;
1037 	struct proc *p;
1038 	int error;
1039 
1040 	options = 0;
1041 	linux_to_bsd_waitopts(args->options, &options);
1042 
1043 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1044 		return (EINVAL);
1045 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1046 		return (EINVAL);
1047 
1048 	switch (args->idtype) {
1049 	case LINUX_P_ALL:
1050 		idtype = P_ALL;
1051 		break;
1052 	case LINUX_P_PID:
1053 		if (args->id <= 0)
1054 			return (EINVAL);
1055 		idtype = P_PID;
1056 		break;
1057 	case LINUX_P_PGID:
1058 		if (args->id <= 0)
1059 			return (EINVAL);
1060 		idtype = P_PGID;
1061 		break;
1062 	default:
1063 		return (EINVAL);
1064 	}
1065 
1066 	error = kern_wait6(td, idtype, args->id, &status, options,
1067 	    &wru, &siginfo);
1068 	if (error != 0)
1069 		return (error);
1070 	if (args->rusage != NULL) {
1071 		error = linux_copyout_rusage(&wru.wru_children,
1072 		    args->rusage);
1073 		if (error != 0)
1074 			return (error);
1075 	}
1076 	if (args->info != NULL) {
1077 		p = td->td_proc;
1078 		bzero(&lsi, sizeof(lsi));
1079 		if (td->td_retval[0] != 0) {
1080 			sig = bsd_to_linux_signal(siginfo.si_signo);
1081 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1082 		}
1083 		error = copyout(&lsi, args->info, sizeof(lsi));
1084 	}
1085 	td->td_retval[0] = 0;
1086 
1087 	return (error);
1088 }
1089 
1090 #ifdef LINUX_LEGACY_SYSCALLS
1091 int
1092 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1093 {
1094 	char *path;
1095 	int error;
1096 	enum uio_seg seg;
1097 	bool convpath;
1098 
1099 	convpath = LUSECONVPATH(td);
1100 	if (!convpath) {
1101 		path = args->path;
1102 		seg = UIO_USERSPACE;
1103 	} else {
1104 		LCONVPATHCREAT(td, args->path, &path);
1105 		seg = UIO_SYSSPACE;
1106 	}
1107 
1108 	switch (args->mode & S_IFMT) {
1109 	case S_IFIFO:
1110 	case S_IFSOCK:
1111 		error = kern_mkfifoat(td, AT_FDCWD, path, seg,
1112 		    args->mode);
1113 		break;
1114 
1115 	case S_IFCHR:
1116 	case S_IFBLK:
1117 		error = kern_mknodat(td, AT_FDCWD, path, seg,
1118 		    args->mode, args->dev);
1119 		break;
1120 
1121 	case S_IFDIR:
1122 		error = EPERM;
1123 		break;
1124 
1125 	case 0:
1126 		args->mode |= S_IFREG;
1127 		/* FALLTHROUGH */
1128 	case S_IFREG:
1129 		error = kern_openat(td, AT_FDCWD, path, seg,
1130 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1131 		if (error == 0)
1132 			kern_close(td, td->td_retval[0]);
1133 		break;
1134 
1135 	default:
1136 		error = EINVAL;
1137 		break;
1138 	}
1139 	if (convpath)
1140 		LFREEPATH(path);
1141 	return (error);
1142 }
1143 #endif
1144 
1145 int
1146 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1147 {
1148 	char *path;
1149 	int error, dfd;
1150 	enum uio_seg seg;
1151 	bool convpath;
1152 
1153 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1154 
1155 	convpath = LUSECONVPATH(td);
1156 	if (!convpath) {
1157 		path = __DECONST(char *, args->filename);
1158 		seg = UIO_USERSPACE;
1159 	} else {
1160 		LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1161 		seg = UIO_SYSSPACE;
1162 	}
1163 
1164 	switch (args->mode & S_IFMT) {
1165 	case S_IFIFO:
1166 	case S_IFSOCK:
1167 		error = kern_mkfifoat(td, dfd, path, seg, args->mode);
1168 		break;
1169 
1170 	case S_IFCHR:
1171 	case S_IFBLK:
1172 		error = kern_mknodat(td, dfd, path, seg, args->mode,
1173 		    args->dev);
1174 		break;
1175 
1176 	case S_IFDIR:
1177 		error = EPERM;
1178 		break;
1179 
1180 	case 0:
1181 		args->mode |= S_IFREG;
1182 		/* FALLTHROUGH */
1183 	case S_IFREG:
1184 		error = kern_openat(td, dfd, path, seg,
1185 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1186 		if (error == 0)
1187 			kern_close(td, td->td_retval[0]);
1188 		break;
1189 
1190 	default:
1191 		error = EINVAL;
1192 		break;
1193 	}
1194 	if (convpath)
1195 		LFREEPATH(path);
1196 	return (error);
1197 }
1198 
1199 /*
1200  * UGH! This is just about the dumbest idea I've ever heard!!
1201  */
1202 int
1203 linux_personality(struct thread *td, struct linux_personality_args *args)
1204 {
1205 	struct linux_pemuldata *pem;
1206 	struct proc *p = td->td_proc;
1207 	uint32_t old;
1208 
1209 	PROC_LOCK(p);
1210 	pem = pem_find(p);
1211 	old = pem->persona;
1212 	if (args->per != 0xffffffff)
1213 		pem->persona = args->per;
1214 	PROC_UNLOCK(p);
1215 
1216 	td->td_retval[0] = old;
1217 	return (0);
1218 }
1219 
1220 struct l_itimerval {
1221 	l_timeval it_interval;
1222 	l_timeval it_value;
1223 };
1224 
1225 #define	B2L_ITIMERVAL(bip, lip)						\
1226 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1227 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1228 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1229 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1230 
1231 int
1232 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1233 {
1234 	int error;
1235 	struct l_itimerval ls;
1236 	struct itimerval aitv, oitv;
1237 
1238 	if (uap->itv == NULL) {
1239 		uap->itv = uap->oitv;
1240 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1241 	}
1242 
1243 	error = copyin(uap->itv, &ls, sizeof(ls));
1244 	if (error != 0)
1245 		return (error);
1246 	B2L_ITIMERVAL(&aitv, &ls);
1247 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1248 	if (error != 0 || uap->oitv == NULL)
1249 		return (error);
1250 	B2L_ITIMERVAL(&ls, &oitv);
1251 
1252 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1253 }
1254 
1255 int
1256 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1257 {
1258 	int error;
1259 	struct l_itimerval ls;
1260 	struct itimerval aitv;
1261 
1262 	error = kern_getitimer(td, uap->which, &aitv);
1263 	if (error != 0)
1264 		return (error);
1265 	B2L_ITIMERVAL(&ls, &aitv);
1266 	return (copyout(&ls, uap->itv, sizeof(ls)));
1267 }
1268 
1269 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1270 int
1271 linux_nice(struct thread *td, struct linux_nice_args *args)
1272 {
1273 
1274 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1275 }
1276 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1277 
1278 int
1279 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1280 {
1281 	struct ucred *newcred, *oldcred;
1282 	l_gid_t *linux_gidset;
1283 	gid_t *bsd_gidset;
1284 	int ngrp, error;
1285 	struct proc *p;
1286 
1287 	ngrp = args->gidsetsize;
1288 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1289 		return (EINVAL);
1290 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1291 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1292 	if (error)
1293 		goto out;
1294 	newcred = crget();
1295 	crextend(newcred, ngrp + 1);
1296 	p = td->td_proc;
1297 	PROC_LOCK(p);
1298 	oldcred = p->p_ucred;
1299 	crcopy(newcred, oldcred);
1300 
1301 	/*
1302 	 * cr_groups[0] holds egid. Setting the whole set from
1303 	 * the supplied set will cause egid to be changed too.
1304 	 * Keep cr_groups[0] unchanged to prevent that.
1305 	 */
1306 
1307 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1308 		PROC_UNLOCK(p);
1309 		crfree(newcred);
1310 		goto out;
1311 	}
1312 
1313 	if (ngrp > 0) {
1314 		newcred->cr_ngroups = ngrp + 1;
1315 
1316 		bsd_gidset = newcred->cr_groups;
1317 		ngrp--;
1318 		while (ngrp >= 0) {
1319 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1320 			ngrp--;
1321 		}
1322 	} else
1323 		newcred->cr_ngroups = 1;
1324 
1325 	setsugid(p);
1326 	proc_set_cred(p, newcred);
1327 	PROC_UNLOCK(p);
1328 	crfree(oldcred);
1329 	error = 0;
1330 out:
1331 	free(linux_gidset, M_LINUX);
1332 	return (error);
1333 }
1334 
1335 int
1336 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1337 {
1338 	struct ucred *cred;
1339 	l_gid_t *linux_gidset;
1340 	gid_t *bsd_gidset;
1341 	int bsd_gidsetsz, ngrp, error;
1342 
1343 	cred = td->td_ucred;
1344 	bsd_gidset = cred->cr_groups;
1345 	bsd_gidsetsz = cred->cr_ngroups - 1;
1346 
1347 	/*
1348 	 * cr_groups[0] holds egid. Returning the whole set
1349 	 * here will cause a duplicate. Exclude cr_groups[0]
1350 	 * to prevent that.
1351 	 */
1352 
1353 	if ((ngrp = args->gidsetsize) == 0) {
1354 		td->td_retval[0] = bsd_gidsetsz;
1355 		return (0);
1356 	}
1357 
1358 	if (ngrp < bsd_gidsetsz)
1359 		return (EINVAL);
1360 
1361 	ngrp = 0;
1362 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1363 	    M_LINUX, M_WAITOK);
1364 	while (ngrp < bsd_gidsetsz) {
1365 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1366 		ngrp++;
1367 	}
1368 
1369 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1370 	free(linux_gidset, M_LINUX);
1371 	if (error)
1372 		return (error);
1373 
1374 	td->td_retval[0] = ngrp;
1375 	return (0);
1376 }
1377 
1378 int
1379 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1380 {
1381 	struct rlimit bsd_rlim;
1382 	struct l_rlimit rlim;
1383 	u_int which;
1384 	int error;
1385 
1386 	if (args->resource >= LINUX_RLIM_NLIMITS)
1387 		return (EINVAL);
1388 
1389 	which = linux_to_bsd_resource[args->resource];
1390 	if (which == -1)
1391 		return (EINVAL);
1392 
1393 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1394 	if (error)
1395 		return (error);
1396 
1397 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1398 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1399 	return (kern_setrlimit(td, which, &bsd_rlim));
1400 }
1401 
1402 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1403 int
1404 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1405 {
1406 	struct l_rlimit rlim;
1407 	struct rlimit bsd_rlim;
1408 	u_int which;
1409 
1410 	if (args->resource >= LINUX_RLIM_NLIMITS)
1411 		return (EINVAL);
1412 
1413 	which = linux_to_bsd_resource[args->resource];
1414 	if (which == -1)
1415 		return (EINVAL);
1416 
1417 	lim_rlimit(td, which, &bsd_rlim);
1418 
1419 #ifdef COMPAT_LINUX32
1420 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1421 	if (rlim.rlim_cur == UINT_MAX)
1422 		rlim.rlim_cur = INT_MAX;
1423 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1424 	if (rlim.rlim_max == UINT_MAX)
1425 		rlim.rlim_max = INT_MAX;
1426 #else
1427 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1428 	if (rlim.rlim_cur == ULONG_MAX)
1429 		rlim.rlim_cur = LONG_MAX;
1430 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1431 	if (rlim.rlim_max == ULONG_MAX)
1432 		rlim.rlim_max = LONG_MAX;
1433 #endif
1434 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1435 }
1436 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1437 
1438 int
1439 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1440 {
1441 	struct l_rlimit rlim;
1442 	struct rlimit bsd_rlim;
1443 	u_int which;
1444 
1445 	if (args->resource >= LINUX_RLIM_NLIMITS)
1446 		return (EINVAL);
1447 
1448 	which = linux_to_bsd_resource[args->resource];
1449 	if (which == -1)
1450 		return (EINVAL);
1451 
1452 	lim_rlimit(td, which, &bsd_rlim);
1453 
1454 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1455 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1456 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1457 }
1458 
1459 int
1460 linux_sched_setscheduler(struct thread *td,
1461     struct linux_sched_setscheduler_args *args)
1462 {
1463 	struct sched_param sched_param;
1464 	struct thread *tdt;
1465 	int error, policy;
1466 
1467 	switch (args->policy) {
1468 	case LINUX_SCHED_OTHER:
1469 		policy = SCHED_OTHER;
1470 		break;
1471 	case LINUX_SCHED_FIFO:
1472 		policy = SCHED_FIFO;
1473 		break;
1474 	case LINUX_SCHED_RR:
1475 		policy = SCHED_RR;
1476 		break;
1477 	default:
1478 		return (EINVAL);
1479 	}
1480 
1481 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1482 	if (error)
1483 		return (error);
1484 
1485 	if (linux_map_sched_prio) {
1486 		switch (policy) {
1487 		case SCHED_OTHER:
1488 			if (sched_param.sched_priority != 0)
1489 				return (EINVAL);
1490 
1491 			sched_param.sched_priority =
1492 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1493 			break;
1494 		case SCHED_FIFO:
1495 		case SCHED_RR:
1496 			if (sched_param.sched_priority < 1 ||
1497 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1498 				return (EINVAL);
1499 
1500 			/*
1501 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1502 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1503 			 */
1504 			sched_param.sched_priority =
1505 			    (sched_param.sched_priority - 1) *
1506 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1507 			    (LINUX_MAX_RT_PRIO - 1);
1508 			break;
1509 		}
1510 	}
1511 
1512 	tdt = linux_tdfind(td, args->pid, -1);
1513 	if (tdt == NULL)
1514 		return (ESRCH);
1515 
1516 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1517 	PROC_UNLOCK(tdt->td_proc);
1518 	return (error);
1519 }
1520 
1521 int
1522 linux_sched_getscheduler(struct thread *td,
1523     struct linux_sched_getscheduler_args *args)
1524 {
1525 	struct thread *tdt;
1526 	int error, policy;
1527 
1528 	tdt = linux_tdfind(td, args->pid, -1);
1529 	if (tdt == NULL)
1530 		return (ESRCH);
1531 
1532 	error = kern_sched_getscheduler(td, tdt, &policy);
1533 	PROC_UNLOCK(tdt->td_proc);
1534 
1535 	switch (policy) {
1536 	case SCHED_OTHER:
1537 		td->td_retval[0] = LINUX_SCHED_OTHER;
1538 		break;
1539 	case SCHED_FIFO:
1540 		td->td_retval[0] = LINUX_SCHED_FIFO;
1541 		break;
1542 	case SCHED_RR:
1543 		td->td_retval[0] = LINUX_SCHED_RR;
1544 		break;
1545 	}
1546 	return (error);
1547 }
1548 
1549 int
1550 linux_sched_get_priority_max(struct thread *td,
1551     struct linux_sched_get_priority_max_args *args)
1552 {
1553 	struct sched_get_priority_max_args bsd;
1554 
1555 	if (linux_map_sched_prio) {
1556 		switch (args->policy) {
1557 		case LINUX_SCHED_OTHER:
1558 			td->td_retval[0] = 0;
1559 			return (0);
1560 		case LINUX_SCHED_FIFO:
1561 		case LINUX_SCHED_RR:
1562 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1563 			return (0);
1564 		default:
1565 			return (EINVAL);
1566 		}
1567 	}
1568 
1569 	switch (args->policy) {
1570 	case LINUX_SCHED_OTHER:
1571 		bsd.policy = SCHED_OTHER;
1572 		break;
1573 	case LINUX_SCHED_FIFO:
1574 		bsd.policy = SCHED_FIFO;
1575 		break;
1576 	case LINUX_SCHED_RR:
1577 		bsd.policy = SCHED_RR;
1578 		break;
1579 	default:
1580 		return (EINVAL);
1581 	}
1582 	return (sys_sched_get_priority_max(td, &bsd));
1583 }
1584 
1585 int
1586 linux_sched_get_priority_min(struct thread *td,
1587     struct linux_sched_get_priority_min_args *args)
1588 {
1589 	struct sched_get_priority_min_args bsd;
1590 
1591 	if (linux_map_sched_prio) {
1592 		switch (args->policy) {
1593 		case LINUX_SCHED_OTHER:
1594 			td->td_retval[0] = 0;
1595 			return (0);
1596 		case LINUX_SCHED_FIFO:
1597 		case LINUX_SCHED_RR:
1598 			td->td_retval[0] = 1;
1599 			return (0);
1600 		default:
1601 			return (EINVAL);
1602 		}
1603 	}
1604 
1605 	switch (args->policy) {
1606 	case LINUX_SCHED_OTHER:
1607 		bsd.policy = SCHED_OTHER;
1608 		break;
1609 	case LINUX_SCHED_FIFO:
1610 		bsd.policy = SCHED_FIFO;
1611 		break;
1612 	case LINUX_SCHED_RR:
1613 		bsd.policy = SCHED_RR;
1614 		break;
1615 	default:
1616 		return (EINVAL);
1617 	}
1618 	return (sys_sched_get_priority_min(td, &bsd));
1619 }
1620 
1621 #define REBOOT_CAD_ON	0x89abcdef
1622 #define REBOOT_CAD_OFF	0
1623 #define REBOOT_HALT	0xcdef0123
1624 #define REBOOT_RESTART	0x01234567
1625 #define REBOOT_RESTART2	0xA1B2C3D4
1626 #define REBOOT_POWEROFF	0x4321FEDC
1627 #define REBOOT_MAGIC1	0xfee1dead
1628 #define REBOOT_MAGIC2	0x28121969
1629 #define REBOOT_MAGIC2A	0x05121996
1630 #define REBOOT_MAGIC2B	0x16041998
1631 
1632 int
1633 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1634 {
1635 	struct reboot_args bsd_args;
1636 
1637 	if (args->magic1 != REBOOT_MAGIC1)
1638 		return (EINVAL);
1639 
1640 	switch (args->magic2) {
1641 	case REBOOT_MAGIC2:
1642 	case REBOOT_MAGIC2A:
1643 	case REBOOT_MAGIC2B:
1644 		break;
1645 	default:
1646 		return (EINVAL);
1647 	}
1648 
1649 	switch (args->cmd) {
1650 	case REBOOT_CAD_ON:
1651 	case REBOOT_CAD_OFF:
1652 		return (priv_check(td, PRIV_REBOOT));
1653 	case REBOOT_HALT:
1654 		bsd_args.opt = RB_HALT;
1655 		break;
1656 	case REBOOT_RESTART:
1657 	case REBOOT_RESTART2:
1658 		bsd_args.opt = 0;
1659 		break;
1660 	case REBOOT_POWEROFF:
1661 		bsd_args.opt = RB_POWEROFF;
1662 		break;
1663 	default:
1664 		return (EINVAL);
1665 	}
1666 	return (sys_reboot(td, &bsd_args));
1667 }
1668 
1669 
1670 int
1671 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1672 {
1673 
1674 	td->td_retval[0] = td->td_proc->p_pid;
1675 
1676 	return (0);
1677 }
1678 
1679 int
1680 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1681 {
1682 	struct linux_emuldata *em;
1683 
1684 	em = em_find(td);
1685 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1686 
1687 	td->td_retval[0] = em->em_tid;
1688 
1689 	return (0);
1690 }
1691 
1692 
1693 int
1694 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1695 {
1696 
1697 	td->td_retval[0] = kern_getppid(td);
1698 	return (0);
1699 }
1700 
1701 int
1702 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1703 {
1704 
1705 	td->td_retval[0] = td->td_ucred->cr_rgid;
1706 	return (0);
1707 }
1708 
1709 int
1710 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1711 {
1712 
1713 	td->td_retval[0] = td->td_ucred->cr_ruid;
1714 	return (0);
1715 }
1716 
1717 int
1718 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1719 {
1720 
1721 	return (kern_getsid(td, args->pid));
1722 }
1723 
1724 int
1725 linux_nosys(struct thread *td, struct nosys_args *ignore)
1726 {
1727 
1728 	return (ENOSYS);
1729 }
1730 
1731 int
1732 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1733 {
1734 	int error;
1735 
1736 	error = kern_getpriority(td, args->which, args->who);
1737 	td->td_retval[0] = 20 - td->td_retval[0];
1738 	return (error);
1739 }
1740 
1741 int
1742 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1743 {
1744 	int name[2];
1745 
1746 	name[0] = CTL_KERN;
1747 	name[1] = KERN_HOSTNAME;
1748 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1749 	    args->len, 0, 0));
1750 }
1751 
1752 int
1753 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1754 {
1755 	int name[2];
1756 
1757 	name[0] = CTL_KERN;
1758 	name[1] = KERN_NISDOMAINNAME;
1759 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1760 	    args->len, 0, 0));
1761 }
1762 
1763 int
1764 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1765 {
1766 
1767 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1768 	    args->error_code);
1769 
1770 	/*
1771 	 * XXX: we should send a signal to the parent if
1772 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1773 	 * as it doesnt occur often.
1774 	 */
1775 	exit1(td, args->error_code, 0);
1776 		/* NOTREACHED */
1777 }
1778 
1779 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1780 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1781 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1782 
1783 struct l_user_cap_header {
1784 	l_int	version;
1785 	l_int	pid;
1786 };
1787 
1788 struct l_user_cap_data {
1789 	l_int	effective;
1790 	l_int	permitted;
1791 	l_int	inheritable;
1792 };
1793 
1794 int
1795 linux_capget(struct thread *td, struct linux_capget_args *uap)
1796 {
1797 	struct l_user_cap_header luch;
1798 	struct l_user_cap_data lucd[2];
1799 	int error, u32s;
1800 
1801 	if (uap->hdrp == NULL)
1802 		return (EFAULT);
1803 
1804 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1805 	if (error != 0)
1806 		return (error);
1807 
1808 	switch (luch.version) {
1809 	case _LINUX_CAPABILITY_VERSION_1:
1810 		u32s = 1;
1811 		break;
1812 	case _LINUX_CAPABILITY_VERSION_2:
1813 	case _LINUX_CAPABILITY_VERSION_3:
1814 		u32s = 2;
1815 		break;
1816 	default:
1817 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1818 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1819 		if (error)
1820 			return (error);
1821 		return (EINVAL);
1822 	}
1823 
1824 	if (luch.pid)
1825 		return (EPERM);
1826 
1827 	if (uap->datap) {
1828 		/*
1829 		 * The current implementation doesn't support setting
1830 		 * a capability (it's essentially a stub) so indicate
1831 		 * that no capabilities are currently set or available
1832 		 * to request.
1833 		 */
1834 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1835 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1836 	}
1837 
1838 	return (error);
1839 }
1840 
1841 int
1842 linux_capset(struct thread *td, struct linux_capset_args *uap)
1843 {
1844 	struct l_user_cap_header luch;
1845 	struct l_user_cap_data lucd[2];
1846 	int error, i, u32s;
1847 
1848 	if (uap->hdrp == NULL || uap->datap == NULL)
1849 		return (EFAULT);
1850 
1851 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1852 	if (error != 0)
1853 		return (error);
1854 
1855 	switch (luch.version) {
1856 	case _LINUX_CAPABILITY_VERSION_1:
1857 		u32s = 1;
1858 		break;
1859 	case _LINUX_CAPABILITY_VERSION_2:
1860 	case _LINUX_CAPABILITY_VERSION_3:
1861 		u32s = 2;
1862 		break;
1863 	default:
1864 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1865 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1866 		if (error)
1867 			return (error);
1868 		return (EINVAL);
1869 	}
1870 
1871 	if (luch.pid)
1872 		return (EPERM);
1873 
1874 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1875 	if (error != 0)
1876 		return (error);
1877 
1878 	/* We currently don't support setting any capabilities. */
1879 	for (i = 0; i < u32s; i++) {
1880 		if (lucd[i].effective || lucd[i].permitted ||
1881 		    lucd[i].inheritable) {
1882 			linux_msg(td,
1883 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1884 			    "inheritable=0x%x is not implemented", i,
1885 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1886 			    (int)lucd[i].inheritable);
1887 			return (EPERM);
1888 		}
1889 	}
1890 
1891 	return (0);
1892 }
1893 
1894 int
1895 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1896 {
1897 	int error = 0, max_size;
1898 	struct proc *p = td->td_proc;
1899 	char comm[LINUX_MAX_COMM_LEN];
1900 	int pdeath_signal;
1901 
1902 	switch (args->option) {
1903 	case LINUX_PR_SET_PDEATHSIG:
1904 		if (!LINUX_SIG_VALID(args->arg2))
1905 			return (EINVAL);
1906 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1907 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1908 		    &pdeath_signal));
1909 	case LINUX_PR_GET_PDEATHSIG:
1910 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1911 		    &pdeath_signal);
1912 		if (error != 0)
1913 			return (error);
1914 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1915 		return (copyout(&pdeath_signal,
1916 		    (void *)(register_t)args->arg2,
1917 		    sizeof(pdeath_signal)));
1918 		break;
1919 	case LINUX_PR_GET_KEEPCAPS:
1920 		/*
1921 		 * Indicate that we always clear the effective and
1922 		 * permitted capability sets when the user id becomes
1923 		 * non-zero (actually the capability sets are simply
1924 		 * always zero in the current implementation).
1925 		 */
1926 		td->td_retval[0] = 0;
1927 		break;
1928 	case LINUX_PR_SET_KEEPCAPS:
1929 		/*
1930 		 * Ignore requests to keep the effective and permitted
1931 		 * capability sets when the user id becomes non-zero.
1932 		 */
1933 		break;
1934 	case LINUX_PR_SET_NAME:
1935 		/*
1936 		 * To be on the safe side we need to make sure to not
1937 		 * overflow the size a Linux program expects. We already
1938 		 * do this here in the copyin, so that we don't need to
1939 		 * check on copyout.
1940 		 */
1941 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1942 		error = copyinstr((void *)(register_t)args->arg2, comm,
1943 		    max_size, NULL);
1944 
1945 		/* Linux silently truncates the name if it is too long. */
1946 		if (error == ENAMETOOLONG) {
1947 			/*
1948 			 * XXX: copyinstr() isn't documented to populate the
1949 			 * array completely, so do a copyin() to be on the
1950 			 * safe side. This should be changed in case
1951 			 * copyinstr() is changed to guarantee this.
1952 			 */
1953 			error = copyin((void *)(register_t)args->arg2, comm,
1954 			    max_size - 1);
1955 			comm[max_size - 1] = '\0';
1956 		}
1957 		if (error)
1958 			return (error);
1959 
1960 		PROC_LOCK(p);
1961 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1962 		PROC_UNLOCK(p);
1963 		break;
1964 	case LINUX_PR_GET_NAME:
1965 		PROC_LOCK(p);
1966 		strlcpy(comm, p->p_comm, sizeof(comm));
1967 		PROC_UNLOCK(p);
1968 		error = copyout(comm, (void *)(register_t)args->arg2,
1969 		    strlen(comm) + 1);
1970 		break;
1971 	default:
1972 		error = EINVAL;
1973 		break;
1974 	}
1975 
1976 	return (error);
1977 }
1978 
1979 int
1980 linux_sched_setparam(struct thread *td,
1981     struct linux_sched_setparam_args *uap)
1982 {
1983 	struct sched_param sched_param;
1984 	struct thread *tdt;
1985 	int error, policy;
1986 
1987 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1988 	if (error)
1989 		return (error);
1990 
1991 	tdt = linux_tdfind(td, uap->pid, -1);
1992 	if (tdt == NULL)
1993 		return (ESRCH);
1994 
1995 	if (linux_map_sched_prio) {
1996 		error = kern_sched_getscheduler(td, tdt, &policy);
1997 		if (error)
1998 			goto out;
1999 
2000 		switch (policy) {
2001 		case SCHED_OTHER:
2002 			if (sched_param.sched_priority != 0) {
2003 				error = EINVAL;
2004 				goto out;
2005 			}
2006 			sched_param.sched_priority =
2007 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
2008 			break;
2009 		case SCHED_FIFO:
2010 		case SCHED_RR:
2011 			if (sched_param.sched_priority < 1 ||
2012 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
2013 				error = EINVAL;
2014 				goto out;
2015 			}
2016 			/*
2017 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
2018 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
2019 			 */
2020 			sched_param.sched_priority =
2021 			    (sched_param.sched_priority - 1) *
2022 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
2023 			    (LINUX_MAX_RT_PRIO - 1);
2024 			break;
2025 		}
2026 	}
2027 
2028 	error = kern_sched_setparam(td, tdt, &sched_param);
2029 out:	PROC_UNLOCK(tdt->td_proc);
2030 	return (error);
2031 }
2032 
2033 int
2034 linux_sched_getparam(struct thread *td,
2035     struct linux_sched_getparam_args *uap)
2036 {
2037 	struct sched_param sched_param;
2038 	struct thread *tdt;
2039 	int error, policy;
2040 
2041 	tdt = linux_tdfind(td, uap->pid, -1);
2042 	if (tdt == NULL)
2043 		return (ESRCH);
2044 
2045 	error = kern_sched_getparam(td, tdt, &sched_param);
2046 	if (error) {
2047 		PROC_UNLOCK(tdt->td_proc);
2048 		return (error);
2049 	}
2050 
2051 	if (linux_map_sched_prio) {
2052 		error = kern_sched_getscheduler(td, tdt, &policy);
2053 		PROC_UNLOCK(tdt->td_proc);
2054 		if (error)
2055 			return (error);
2056 
2057 		switch (policy) {
2058 		case SCHED_OTHER:
2059 			sched_param.sched_priority = 0;
2060 			break;
2061 		case SCHED_FIFO:
2062 		case SCHED_RR:
2063 			/*
2064 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2065 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2066 			 */
2067 			sched_param.sched_priority =
2068 			    (sched_param.sched_priority *
2069 			    (LINUX_MAX_RT_PRIO - 1) +
2070 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2071 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2072 			break;
2073 		}
2074 	} else
2075 		PROC_UNLOCK(tdt->td_proc);
2076 
2077 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
2078 	return (error);
2079 }
2080 
2081 /*
2082  * Get affinity of a process.
2083  */
2084 int
2085 linux_sched_getaffinity(struct thread *td,
2086     struct linux_sched_getaffinity_args *args)
2087 {
2088 	int error;
2089 	struct thread *tdt;
2090 
2091 	if (args->len < sizeof(cpuset_t))
2092 		return (EINVAL);
2093 
2094 	tdt = linux_tdfind(td, args->pid, -1);
2095 	if (tdt == NULL)
2096 		return (ESRCH);
2097 
2098 	PROC_UNLOCK(tdt->td_proc);
2099 
2100 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2101 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2102 	if (error == 0)
2103 		td->td_retval[0] = sizeof(cpuset_t);
2104 
2105 	return (error);
2106 }
2107 
2108 /*
2109  *  Set affinity of a process.
2110  */
2111 int
2112 linux_sched_setaffinity(struct thread *td,
2113     struct linux_sched_setaffinity_args *args)
2114 {
2115 	struct thread *tdt;
2116 
2117 	if (args->len < sizeof(cpuset_t))
2118 		return (EINVAL);
2119 
2120 	tdt = linux_tdfind(td, args->pid, -1);
2121 	if (tdt == NULL)
2122 		return (ESRCH);
2123 
2124 	PROC_UNLOCK(tdt->td_proc);
2125 
2126 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2127 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2128 }
2129 
2130 struct linux_rlimit64 {
2131 	uint64_t	rlim_cur;
2132 	uint64_t	rlim_max;
2133 };
2134 
2135 int
2136 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2137 {
2138 	struct rlimit rlim, nrlim;
2139 	struct linux_rlimit64 lrlim;
2140 	struct proc *p;
2141 	u_int which;
2142 	int flags;
2143 	int error;
2144 
2145 	if (args->resource >= LINUX_RLIM_NLIMITS)
2146 		return (EINVAL);
2147 
2148 	which = linux_to_bsd_resource[args->resource];
2149 	if (which == -1)
2150 		return (EINVAL);
2151 
2152 	if (args->new != NULL) {
2153 		/*
2154 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2155 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2156 		 * as INFINITY so we do not need a conversion even.
2157 		 */
2158 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2159 		if (error != 0)
2160 			return (error);
2161 	}
2162 
2163 	flags = PGET_HOLD | PGET_NOTWEXIT;
2164 	if (args->new != NULL)
2165 		flags |= PGET_CANDEBUG;
2166 	else
2167 		flags |= PGET_CANSEE;
2168 	if (args->pid == 0) {
2169 		p = td->td_proc;
2170 		PHOLD(p);
2171 	} else {
2172 		error = pget(args->pid, flags, &p);
2173 		if (error != 0)
2174 			return (error);
2175 	}
2176 	if (args->old != NULL) {
2177 		PROC_LOCK(p);
2178 		lim_rlimit_proc(p, which, &rlim);
2179 		PROC_UNLOCK(p);
2180 		if (rlim.rlim_cur == RLIM_INFINITY)
2181 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2182 		else
2183 			lrlim.rlim_cur = rlim.rlim_cur;
2184 		if (rlim.rlim_max == RLIM_INFINITY)
2185 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2186 		else
2187 			lrlim.rlim_max = rlim.rlim_max;
2188 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2189 		if (error != 0)
2190 			goto out;
2191 	}
2192 
2193 	if (args->new != NULL)
2194 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2195 
2196  out:
2197 	PRELE(p);
2198 	return (error);
2199 }
2200 
2201 int
2202 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2203 {
2204 	struct timeval utv, tv0, tv1, *tvp;
2205 	struct l_pselect6arg lpse6;
2206 	struct l_timespec lts;
2207 	struct timespec uts;
2208 	l_sigset_t l_ss;
2209 	sigset_t *ssp;
2210 	sigset_t ss;
2211 	int error;
2212 
2213 	ssp = NULL;
2214 	if (args->sig != NULL) {
2215 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2216 		if (error != 0)
2217 			return (error);
2218 		if (lpse6.ss_len != sizeof(l_ss))
2219 			return (EINVAL);
2220 		if (lpse6.ss != 0) {
2221 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2222 			    sizeof(l_ss));
2223 			if (error != 0)
2224 				return (error);
2225 			linux_to_bsd_sigset(&l_ss, &ss);
2226 			ssp = &ss;
2227 		}
2228 	}
2229 
2230 	/*
2231 	 * Currently glibc changes nanosecond number to microsecond.
2232 	 * This mean losing precision but for now it is hardly seen.
2233 	 */
2234 	if (args->tsp != NULL) {
2235 		error = copyin(args->tsp, &lts, sizeof(lts));
2236 		if (error != 0)
2237 			return (error);
2238 		error = linux_to_native_timespec(&uts, &lts);
2239 		if (error != 0)
2240 			return (error);
2241 
2242 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2243 		if (itimerfix(&utv))
2244 			return (EINVAL);
2245 
2246 		microtime(&tv0);
2247 		tvp = &utv;
2248 	} else
2249 		tvp = NULL;
2250 
2251 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2252 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2253 
2254 	if (error == 0 && args->tsp != NULL) {
2255 		if (td->td_retval[0] != 0) {
2256 			/*
2257 			 * Compute how much time was left of the timeout,
2258 			 * by subtracting the current time and the time
2259 			 * before we started the call, and subtracting
2260 			 * that result from the user-supplied value.
2261 			 */
2262 
2263 			microtime(&tv1);
2264 			timevalsub(&tv1, &tv0);
2265 			timevalsub(&utv, &tv1);
2266 			if (utv.tv_sec < 0)
2267 				timevalclear(&utv);
2268 		} else
2269 			timevalclear(&utv);
2270 
2271 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2272 
2273 		error = native_to_linux_timespec(&lts, &uts);
2274 		if (error == 0)
2275 			error = copyout(&lts, args->tsp, sizeof(lts));
2276 	}
2277 
2278 	return (error);
2279 }
2280 
2281 int
2282 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2283 {
2284 	struct timespec ts0, ts1;
2285 	struct l_timespec lts;
2286 	struct timespec uts, *tsp;
2287 	l_sigset_t l_ss;
2288 	sigset_t *ssp;
2289 	sigset_t ss;
2290 	int error;
2291 
2292 	if (args->sset != NULL) {
2293 		if (args->ssize != sizeof(l_ss))
2294 			return (EINVAL);
2295 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2296 		if (error)
2297 			return (error);
2298 		linux_to_bsd_sigset(&l_ss, &ss);
2299 		ssp = &ss;
2300 	} else
2301 		ssp = NULL;
2302 	if (args->tsp != NULL) {
2303 		error = copyin(args->tsp, &lts, sizeof(lts));
2304 		if (error)
2305 			return (error);
2306 		error = linux_to_native_timespec(&uts, &lts);
2307 		if (error != 0)
2308 			return (error);
2309 
2310 		nanotime(&ts0);
2311 		tsp = &uts;
2312 	} else
2313 		tsp = NULL;
2314 
2315 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2316 
2317 	if (error == 0 && args->tsp != NULL) {
2318 		if (td->td_retval[0]) {
2319 			nanotime(&ts1);
2320 			timespecsub(&ts1, &ts0, &ts1);
2321 			timespecsub(&uts, &ts1, &uts);
2322 			if (uts.tv_sec < 0)
2323 				timespecclear(&uts);
2324 		} else
2325 			timespecclear(&uts);
2326 
2327 		error = native_to_linux_timespec(&lts, &uts);
2328 		if (error == 0)
2329 			error = copyout(&lts, args->tsp, sizeof(lts));
2330 	}
2331 
2332 	return (error);
2333 }
2334 
2335 int
2336 linux_sched_rr_get_interval(struct thread *td,
2337     struct linux_sched_rr_get_interval_args *uap)
2338 {
2339 	struct timespec ts;
2340 	struct l_timespec lts;
2341 	struct thread *tdt;
2342 	int error;
2343 
2344 	/*
2345 	 * According to man in case the invalid pid specified
2346 	 * EINVAL should be returned.
2347 	 */
2348 	if (uap->pid < 0)
2349 		return (EINVAL);
2350 
2351 	tdt = linux_tdfind(td, uap->pid, -1);
2352 	if (tdt == NULL)
2353 		return (ESRCH);
2354 
2355 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2356 	PROC_UNLOCK(tdt->td_proc);
2357 	if (error != 0)
2358 		return (error);
2359 	error = native_to_linux_timespec(&lts, &ts);
2360 	if (error != 0)
2361 		return (error);
2362 	return (copyout(&lts, uap->interval, sizeof(lts)));
2363 }
2364 
2365 /*
2366  * In case when the Linux thread is the initial thread in
2367  * the thread group thread id is equal to the process id.
2368  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2369  */
2370 struct thread *
2371 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2372 {
2373 	struct linux_emuldata *em;
2374 	struct thread *tdt;
2375 	struct proc *p;
2376 
2377 	tdt = NULL;
2378 	if (tid == 0 || tid == td->td_tid) {
2379 		tdt = td;
2380 		PROC_LOCK(tdt->td_proc);
2381 	} else if (tid > PID_MAX)
2382 		tdt = tdfind(tid, pid);
2383 	else {
2384 		/*
2385 		 * Initial thread where the tid equal to the pid.
2386 		 */
2387 		p = pfind(tid);
2388 		if (p != NULL) {
2389 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2390 				/*
2391 				 * p is not a Linuxulator process.
2392 				 */
2393 				PROC_UNLOCK(p);
2394 				return (NULL);
2395 			}
2396 			FOREACH_THREAD_IN_PROC(p, tdt) {
2397 				em = em_find(tdt);
2398 				if (tid == em->em_tid)
2399 					return (tdt);
2400 			}
2401 			PROC_UNLOCK(p);
2402 		}
2403 		return (NULL);
2404 	}
2405 
2406 	return (tdt);
2407 }
2408 
2409 void
2410 linux_to_bsd_waitopts(int options, int *bsdopts)
2411 {
2412 
2413 	if (options & LINUX_WNOHANG)
2414 		*bsdopts |= WNOHANG;
2415 	if (options & LINUX_WUNTRACED)
2416 		*bsdopts |= WUNTRACED;
2417 	if (options & LINUX_WEXITED)
2418 		*bsdopts |= WEXITED;
2419 	if (options & LINUX_WCONTINUED)
2420 		*bsdopts |= WCONTINUED;
2421 	if (options & LINUX_WNOWAIT)
2422 		*bsdopts |= WNOWAIT;
2423 
2424 	if (options & __WCLONE)
2425 		*bsdopts |= WLINUXCLONE;
2426 }
2427 
2428 int
2429 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2430 {
2431 	struct uio uio;
2432 	struct iovec iov;
2433 	int error;
2434 
2435 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2436 		return (EINVAL);
2437 	if (args->count > INT_MAX)
2438 		args->count = INT_MAX;
2439 
2440 	iov.iov_base = args->buf;
2441 	iov.iov_len = args->count;
2442 
2443 	uio.uio_iov = &iov;
2444 	uio.uio_iovcnt = 1;
2445 	uio.uio_resid = iov.iov_len;
2446 	uio.uio_segflg = UIO_USERSPACE;
2447 	uio.uio_rw = UIO_READ;
2448 	uio.uio_td = td;
2449 
2450 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2451 	if (error == 0)
2452 		td->td_retval[0] = args->count - uio.uio_resid;
2453 	return (error);
2454 }
2455 
2456 int
2457 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2458 {
2459 
2460 	/* Needs to be page-aligned */
2461 	if (args->start & PAGE_MASK)
2462 		return (EINVAL);
2463 	return (kern_mincore(td, args->start, args->len, args->vec));
2464 }
2465 
2466 #define	SYSLOG_TAG	"<6>"
2467 
2468 int
2469 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2470 {
2471 	char buf[128], *src, *dst;
2472 	u_int seq;
2473 	int buflen, error;
2474 
2475 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2476 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2477 		return (EINVAL);
2478 	}
2479 
2480 	if (args->len < 6) {
2481 		td->td_retval[0] = 0;
2482 		return (0);
2483 	}
2484 
2485 	error = priv_check(td, PRIV_MSGBUF);
2486 	if (error)
2487 		return (error);
2488 
2489 	mtx_lock(&msgbuf_lock);
2490 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2491 	mtx_unlock(&msgbuf_lock);
2492 
2493 	dst = args->buf;
2494 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2495 	/* The -1 is to skip the trailing '\0'. */
2496 	dst += sizeof(SYSLOG_TAG) - 1;
2497 
2498 	while (error == 0) {
2499 		mtx_lock(&msgbuf_lock);
2500 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2501 		mtx_unlock(&msgbuf_lock);
2502 
2503 		if (buflen == 0)
2504 			break;
2505 
2506 		for (src = buf; src < buf + buflen && error == 0; src++) {
2507 			if (*src == '\0')
2508 				continue;
2509 
2510 			if (dst >= args->buf + args->len)
2511 				goto out;
2512 
2513 			error = copyout(src, dst, 1);
2514 			dst++;
2515 
2516 			if (*src == '\n' && *(src + 1) != '<' &&
2517 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2518 				error = copyout(&SYSLOG_TAG,
2519 				    dst, sizeof(SYSLOG_TAG));
2520 				dst += sizeof(SYSLOG_TAG) - 1;
2521 			}
2522 		}
2523 	}
2524 out:
2525 	td->td_retval[0] = dst - args->buf;
2526 	return (error);
2527 }
2528 
2529 int
2530 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2531 {
2532 	int cpu, error, node;
2533 
2534 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2535 	error = 0;
2536 	node = cpuid_to_pcpu[cpu]->pc_domain;
2537 
2538 	if (args->cpu != NULL)
2539 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2540 	if (args->node != NULL)
2541 		error = copyout(&node, args->node, sizeof(l_int));
2542 	return (error);
2543 }
2544